Migrate away from this thin ssh wrapper module while adding more test
coverage.
Removes `sessionPartial`, which hand-copied the layout of a private
`gliderlabs/ssh` struct and reinterpreted a pointer to it via
`reflect.UnsafePointer` to reach the permissions of the authenticated
connection. The layout is unchecked, so an upstream field reorder would
mismatch silently.
The builtin server only needs the session channel with `exec` and
`shell`. Serving those on `x/crypto` drops the hack and the dependency,
since `PublicKeyCallback` returns permissions per key and `x/crypto`
assigns them only after verifying the signature.
Two benign behavior changes:
1. Internal session handler errors report exit status 1 rather than 0,
so a client no longer reads a failure as success.
1. An unusable host key is fatal at startup instead of being replaced by
an ephemeral one that would trigger an error at the client.
Follow-up to https://github.com/go-gitea/gitea/pull/37571.
"Participating and mentions" deleted the watch row, so choosing it
dropped you out of the watcher count. It is a watch like the others, so
it now keeps a row and simply subscribes to no events.
The dashboard feed ignored the per-event options, so a "Custom: issues"
watcher still got pull request activity there. It now gates on the same
options as mail and notifications. That also closes a gap where pull
request reviews bypassed the permission check.
Also, address
https://github.com/go-gitea/gitea/pull/37571#discussion_r3740487363 and
reword a UI text for clarity.
---------
Signed-off-by: silverwind <me@silverwind.io>
Co-authored-by: wxiaoguang <wxiaoguang@gmail.com>
Enable [`forcetypeassert`](https://github.com/gostaticanalysis/forcetypeassert)
linter to prevent unchecked type assertions. ~650 issues fixed, most
fixes were clean, some use `setting.PanicInDevOrTesting`.
The only behaviour changes are where code would previously send a 500 error
or panic, a 4xx error is now emitted.
Co-authored-by: wxiaoguang <wxiaoguang@gmail.com>
Renovate installs a toolchain only for the managers a branch actually
updates. A branch without a gomod update therefore has no `go` on
`PATH`, and `make tidy`
[fails](https://github.com/go-gitea/gitea/pull/38813#issuecomment-5211013198).
Fix by installing those tools. `node` is not needed because it comes
with renovate.
---------
Signed-off-by: silverwind <me@silverwind.io>
Few small tweaks that should improve the quality:
- Make agents read `docs` folder
- Shorter comments, shorter PR descriptions
- Modern Go features
Signed-off-by: silverwind <me@silverwind.io>
The npm packument left `time`, `keywords` and `maintainers` empty
although the data was available. `created` and `modified` are derived
from the versions currently served, as there is no package-level
timestamp to read them from.
Fixes: https://github.com/go-gitea/gitea/issues/38818
Accepts GitHub's `$/` self-repository prefix in a reusable workflow
`uses:`, alongside `./`.
Gitea's `./` already resolves against the caller's own source repo and
commit, which is what `$/` means, so the two are aliases here. Cycle
detection folds both prefixes onto one key.
Related PR for step-level support:
https://gitea.com/gitea/runner/pulls/1150
Removes what would block a later switch to Vue's vapor mode:
`vue-chartjs` and the `SvgIcon` render function are virtual DOM
components, and `v-memo` has no vapor equivalent. This does not adopt
vapor mode, which will be stable in upcoming Vue 3.6.
`vue-chartjs` was a thin wrapper over chart.js, so a local
`ChartCanvas.vue` replaces it. Chart data and options move into computed
values to keep their object identity, which is what `v-memo` was
compensating for.
`chartjs-adapter-dayjs-4` is moved first-party, just ~40 lines that are
easy to maintain.
- The pub registry reported the oldest version as `latest`, because the
descriptor slice is sorted ascending but the first element was used.
- Verifying a GPG or SSH key flashed success and redirected after
already writing an error response, so a failure was reported as a
success with an empty key id.
- Test packages sharing redis could tear down each other's server.
`PrepareTestRedis` started its own on the well-known port, so a package
running in parallel borrowed it and lost it when the owner's cleanup
fired. It now listens on a socket of its own.
Registration omitted `userVerification`, so Chromium raised the
credential to credProtect level 3 and the authenticator then hid it from
the second-factor login, which asked for `discouraged`. Registration and
each login now set their own value, with `preferred` on the second
factor so credentials already registered at level 3 keep working without
re-enrollment.
Also add relevant e2e test coverage for webauthn, one test chromium only
because Firefox lacks the APIs needed.
Fixes https://github.com/go-gitea/gitea/issues/33531
Fixes https://github.com/go-gitea/gitea/issues/36019
Fixes https://github.com/go-gitea/gitea/issues/38139
Every `${{ }}` part was spliced as raw text into a synthesized
`format('...', <raw>)` call and re-parsed, so unbalanced parentheses
restructured the whole expression:
```yaml
run-name: ${{ 1) && (2 }} # panicked, aborting workflow parsing for the push
if: x ${{ 1) && (2 }} y # silently evaluated to 2
```
One scanner shaped like GitHub's template reader now splits every value
and each part is evaluated on its own, so nothing builds an expression
out of text. A part that fails is an error instead of an empty string,
making a `run-name`, `runs-on` or job name that GitHub rejects invalid
here too.
Replaces https://github.com/go-gitea/gitea/pull/38736 as the actual root-cause fix.
Signed-off-by: silverwind <me@silverwind.io>
`UpdateTaskByState` wrote the task, its job and its steps in separate
statements. An interruption in between left the task finished with a
running job, so the run stayed in progress, and the "state is final"
early return made every retry, cancel and cleanup a no-op.
Fixes https://github.com/go-gitea/gitea/issues/38790
Migrates the last four Options API components, so all 20 `.vue` files
now use `<script setup>`, and disables the Vue Options API runtime. This
will enable optimizations like Vue Vapor mode in the future.
Bug fixes done while migrating and testing:
- Branch selector: arrowing onto "Create branch …" threw a `TypeError`,
and Enter on it did nothing
- Dashboard: the search term was not escaped, so `&` injected query
parameters and `#` truncated the request
- Dashboard: an unknown `repo-search-filter` threw, and names like
`constructor` leaked `Object.prototype` members into the request
- Dashboard: an unknown archived/private filter rendered `function
Object() { [native code] }` as a checkbox tooltip
- Dashboard: removed a dropdown init that always ran against a
not-yet-rendered element
- Branch selector: the `document.body` click listener was never removed
on unmount
- Contributors: the chart plugin relied on an undeclared chart.js field
- Contributors: `contributorsStats` was mutated behind a `shallowRef`,
so future readers would not update
- Removed template attributes and pageData keys that no component had
read for years
---------
Signed-off-by: silverwind <me@silverwind.io>
Replaces the `ansi_up` dependency with first-party code and fixes a
number of bugs in turn.
- Faster rendering, around 7x for plain lines and 3x for colored ones.
- Render many SGR features like hyperlinks, blink, inverse, conceal,
strikethrough, overline, underline styles and underline color, including
`:` sub-parameters, which no longer swallow the codes after them.
- Drop OSC, DCS, SOS, PM and APC with their payload, ending them at BEL,
`ESC \` or the 8-bit ST. A truncated sequence is dropped instead of
corrupting a later line.
- A backspace moves the cursor back a column, so what follows overwrites
it, even across a style change.
- A style inside an OSC 8 label renders instead of leaking, and a
private CSI ending in `m` no longer resets the style.
- Log lines render as DOM nodes, never as markup, and only an `http(s)`
url becomes a link.
- Named colors render as CSS classes, only 24-bit color stays inline.
- Invisible text is now selectable, and the `z-index` workaround is
gone.
Signed-off-by: silverwind <me@silverwind.io>
Co-authored-by: Copilot Autofix powered by AI <175728472+Copilot@users.noreply.github.com>
Co-authored-by: wxiaoguang <wxiaoguang@gmail.com>
The hardcoded cipher suites and curve preferences date from 2021 and no
longer match what Go ships. All four `SSL_*` are now unset by default
which means "use Go's default".
1. TLS 1.3 is now the default instead of TLS 1.2.
2. Post-quantum key exchange (`X25519MLKEM768`) works.
3. Four `ECDHE-*-CBC-SHA` suites become negotiable on TLS 1.2.
- drop `dedupePeerDependents` which is a no-op (produces same lockfile)
- replace `allowBuilds` with `ignoreScripts`, no dependency needs any
builds, on any platform
- disable `confirmModulesPurge`, related to
https://github.com/pnpm/pnpm/issues/11562
- disable `verifyDepsBeforeRun`, `make` already ensures `node_modules`
are up to date, this saves around 150ms per `pnpm exec` and reduces pnpm
spam.
Signed-off-by: silverwind <me@silverwind.io>
Co-authored-by: Claude (Opus 5) <noreply@anthropic.com>
Speed up all go jobs by ~58s by not cleaning the disk when there's at
least 50GB available.
The job was added because some runners only have like 15GB available (I
assume self-hosted), but on regular GHA runners which have about 90GB
free, this job will now always skip.
Co-authored-by: techknowlogick <techknowlogick@gitea.com>
On the issue ref popup, when the user clicks the link while fetch is
loading, the page will briefly render a network error message before
loading the next page. Suppress unnecessary network errors.
Fix minor issue: `cache-prune` logs `jq: error: writing output failed:
Broken pipe` when it has nothing to delete, because the loop stops
reading as soon as it is under the limit while `jq` still has output
pending. Reading from a here-string removes the pipe.
---------
Signed-off-by: silverwind <me@silverwind.io>
Cache saves have been rejected since June because the repo sat above its
10 GB allowance, so every PR run fell back to a cache built with an
older toolchain and rebuilt the backend from scratch.
- go version in the `gobuild` and `golint` keys and `restore-keys`
- seeder triggers on `go.mod` and gained `workflow_dispatch`
- single writer for the `gomod` cache, the two were racing
- new daily `cache-prune` workflow holding the total under a limit
- `cache: false` for `setup-go` in release and cron workflows, it held
815 MB
- add `workflow_dispatch` so the cache workflows can be triggered
on-demand
Assisted-by: Claude Code:claude-opus-5
The api specs were Go templates whose committed form was not a valid
swagger document, so `swagger-validate`, `generate-openapi.go` and
`.spectral.yaml` each worked around it. They are now plain json,
substituted at serve time.
Renaming them off `.tmpl` also stops `make fmt` rewriting them, which
used to bump their mtime and silently skip the next `make
generate-swagger`.
Also enables stricter spectral linting: extends `lint-swagger` to the
OpenAPI 3 spec, turns on `openapi-tags`, `operation-singular-tag` and
`operation-tag-defined`, adds a top-level `tags` array with descriptions
to the swagger input, and drops the redundant `repository` tag from
`POST /user/repos`.
---------
Signed-off-by: silverwind <me@silverwind.io>
Co-authored-by: wxiaoguang <wxiaoguang@gmail.com>
Co-authored-by: Giteabot <teabot@gitea.io>
Further refinements to AGENTS.md:
- consolidate some points
- add point about locales
- rewrite some points for clarity and size
Signed-off-by: silverwind <me@silverwind.io>
Co-authored-by: delvh <dev.lh@web.de>
`.markup del code { text-decoration: inherit }` makes inline code inside
`<del>` paint its own line-through in addition to the one already
propagating from the parent. Since `code` is `font-size: 85%`, the two
land at different heights and render as a doubled strikethrough.
The rule was inherited from primer-markdown, where it was added in
https://github.com/primer/css/commit/762b8b8264aa4c4beed0a7f842f90c142eb2b310
("so that `<del>` or `<a>` has the same effects on `<code>` tags") at a
time when inline `code` was `display: inline-block`, a box that text
decorations do not propagate into. That `display: inline-block` was
removed 11 days later in
https://github.com/primer/css/commit/f1131d5ab618ac41d4097823b511ca0b373b2ee2,
which made the rule redundant, but it survived the squashed import into
primer/css and every copy downstream of it.
No other popular markdown stylesheet carries an equivalent rule, GitHub
still ships it and shows the same doubled line.
Fixes: https://github.com/go-gitea/gitea/issues/34786
`RedisBroker.Subscribe` returns before the server acks `SUBSCRIBE`, so a
publish right after it can be dropped, making
`TestRedisBroker/CrossBroker` fail intermittently on loaded CI runners
([example](https://github.com/go-gitea/gitea/actions/runs/30257188479/job/89948425118)).
Each scenario now uses its own topic and waits for `PUBSUB NUMSUB`
before publishing. `MemoryBroker` registers synchronously and skips the
wait.
The preset we extended only matches a `# renovate:` comment on the line
above the `_VERSION` assignment. Ours trails the value, like the ones in
the `Makefile`, so it never matched and the pin has sat at `43.141.5`
since https://github.com/go-gitea/gitea/pull/37050.
Co-authored-by: Claude (Opus 5) <noreply@anthropic.com>
Eliminate two layout shifts in the menu, one related to non-existant
label on page load and one to `0` value rendering.
Signed-off-by: silverwind <me@silverwind.io>
Co-authored-by: Claude (Opus 4.7) <noreply@anthropic.com>
Co-authored-by: bircni <bircni@icloud.com>
Replace the previous main-branch fetch with data from npm package
[`linguist-languages`](https://github.com/ikatyang-collab/linguist-languages)
and only run generate targets when actually needed. The removed
`fileFilters` were leftovers from the nolyfill removal.
- enable `skip-initialism-name-checks`, so names like `sessionUid` are
allowed
- replace removed `skip-package-name-checks` option with new
`package-naming` with fitting rules for gitea
- drop dead exclusion paths
Fix regression from https://github.com/go-gitea/gitea/pull/38610
`perfect-debounce` allowed only one request in flight; `debounce` does
not, so a slow older request can resolve after a newer one and leave a
stale suggestion menu. Superseded requests are now aborted.
Adds `web_src/js/utils/func.ts` with `debounce` and `throttle`, dropping
`throttle-debounce` and `perfect-debounce` dependencies. Both were
needed before: the former has no promise-returning debounce, which
`TextExpander` requires, and the latter no `throttle`.
Signature follows lodash and es-toolkit: `(func, wait, {leading,
trailing})` plus `cancel`, so argument order flips at the migrated call
sites.
Derives topic labels from the PR title scope: `enhance(actions): ...`
gets `topic/gitea-actions`. Scopes map to `topic/<scope>`, with aliases
for names that differ (`oauth` → `topic/authentication`).
Labels stay in sync with the title, while ones applied by hand are left
alone. Scopes with no matching label on the repo are ignored, so no new
labels get created.
---------
Co-authored-by: Giteabot <teabot@gitea.io>
- avoid layout shift on actions log view page load
- retain header border when actions log is scrolled
- when copying step output, only copy with timestamp when timestamp
display is enabled (useful for example when step output is plain json
and copying with timestamp would make that json unparsable).
Remember the scroll offset when the merge box reloads. This is a crude
patch now, a proper fix would be a fully vue-rendered template with a
json payload from backend, which is much more work. Idiomorph was also
considered but deemed unviable.
Adds browser-level coverage for the pull request merge box, which
currently had no tests. Exercises the full merge flow: open the PR,
expand the merge form, submit, and assert the merged state.
Extracted from https://github.com/go-gitea/gitea/pull/36759.
Also updated AGENTS.md with instructions for e2e tests.
---------
Signed-off-by: silverwind <me@silverwind.io>
Co-authored-by: Giteabot <teabot@gitea.io>
Replace the single-use `uint8-to-base64` dependency with native
`btoa`/`atob`. The implementation is exactly identical.
---------
Signed-off-by: silverwind <me@silverwind.io>
- split the go version in `go.mod` into `go` and `toolchain` again
- add a workaround for https://github.com/golang/go/issues/75331 so
`make tidy` never drops `toolchain`
- configure renovate to bump `go` on minor releases and `toolchain` on
every release
- go and toolchain bumps land in a separate, fast-tracked PR, as both
carry security fixes
Replaces: https://github.com/go-gitea/gitea/pull/37846
Group updates from all managers into a single `dependencies` group so
monday produces one combined PR instead of one per manager, matching
what is regularly done manually to reduce CI waiting time. The `gomod`
and `npm` rules remain for their `postUpgradeTasks`, which run once on
the combined branch.
Also replace the all-day monday schedule with the `schedule:weekly`
preset (monday 0-4 UTC) so updates whose `minimumReleaseAge` lapses
later in the day wait for the next weekly batch instead of raising a new
PR the same day, as happened in
https://github.com/go-gitea/gitea/pull/38546. `vulnerabilityAlerts` and
the go toolchain rule bypass the schedule as before.
The test raced the server-side SSE channel registration: a logout
emitted before registration is silently dropped ([example
flake](https://github.com/go-gitea/gitea/actions/runs/29699349255/job/88225654080)).
The 500ms wait from https://github.com/go-gitea/gitea/pull/37403 only
made this unlikely.
The server now registers the channel before sending the initial response
bytes, so an open connection implies registration. The shared worker
forwards the built-in `open` event (replaying it to late-attaching ports
via `EventSource.readyState`), the page exposes it as a
`data-user-events-connected` attribute, and the test waits for that
attribute instead of a fixed timeout.
- update eslint and related dependencies, `eslint-plugin-unicorn` to v72
with new rules enabled
- remove `eslint-plugin-sonarjs` (mostly obsolete or slow rules)
- extract shared eslint args into `ESLINT_ARGS` make variable
- rewrite ignore patterns in `eslint.json.config.ts` as explicit
root-relative paths: add `data` (local server data failed `lint-json`),
drop `node_modules` (ignored by default)
- fix a broken link in `CONTRIBUTING.md` (discovered during markdown
linting with eslint, but decided to not add that yet)
Co-authored-by: wxiaoguang <wxiaoguang@gmail.com>
Follow-ups from https://github.com/go-gitea/gitea/pull/37038:
- Render the OpenAPI 3.0 spec in the API viewer, it is richer than the
Swagger 2.0 rendering
- Replace the back link with gitea-styled buttons to view both specs and
return to Gitea, flowing above swagger-ui on viewports where they would
overlap the title
- Key `VisibilityModes` by the string enum type, now named
`VisibilityString`, removing the `string()` casts at API call sites
- Drop the raw visibility strings from the `Service` settings struct in
favor of the typed mode fields, which also removes the org default
visibility row from the admin config page
- Fix two pre-existing issues surfaced in review: an invalid
`DEFAULT_USER_VISIBILITY` was silently accepted as public, and the org
visibility error message showed the enum zero value instead of the
submitted input
Co-authored-by: wxiaoguang <wxiaoguang@gmail.com>
1. combine PRs for major/non-major, less PRs is less work with the
fixups.
2. always set `chore`, more often than not, this is more correct then
`fix`.
---------
Signed-off-by: silverwind <me@silverwind.io>
Scheduled runs stored a `null` event payload, so
`github.event.repository`/`sender`/`organization` were empty in
scheduled workflows. Every other event carries them, and so does GitHub
Actions. `CreateScheduleTask` now synthesizes them.
https://github.com/go-gitea/gitea/pull/37594 widened the author column
from `three wide` to `four wide`, leaving a large empty gap around the
SHA column in the common single-author case. The old author cell was
capped at 180px, so the wider column only adds whitespace: avatar-stack
names ellipsize at 240px each, and wider multi-author rows still expand
naturally in the auto-layout table. Restore the pre-existing
`three`/`eight` widths.
Co-authored-by: bircni <bircni@icloud.com>
`data-render-name` is set before the plugin's async render runs, so
measuring the container height right after the attribute appears can
observe the pre-render 48px height when the `pdfobject` chunk loads
slowly (flaked in CI). Poll for the height instead, like the asciicast
test in the same file does.
The nightly snap build fails with `snapcraft remote-build`'s
`BadRequest()`: it force-pushes Gitea's full history to a fresh
Launchpad repo each run and creates the recipe before Launchpad has
indexed the `main` ref. The race is unwinnable at Gitea's repo size, and
Canonical does not support `remote-build` in CI.
Build locally on native amd64 + arm64 runners with
`snapcore/action-build` + `snapcore/action-publish` instead — no
Launchpad, no race. Drops the `LAUNCHPAD_CREDENTIALS` secret (publishing
keeps `SNAPCRAFT_STORE_CREDENTIALS`); the `snap/` recipe is unchanged,
so the same snaps ship to `latest/edge`.
Co-authored-by: Lunny Xiao <xiaolunwen@gmail.com>
Launchpad no longer builds `core24` snaps for `armhf`. Since `snapcraft
remote-build` submits a single request for all platforms, the `armhf`
rejection fails with `snapcraft internal error: BadRequest()` and takes
the `amd64` and `arm64` builds down with it, so nothing gets published.
Dropping 32-bit ARM from `snap/snapcraft.yaml` and the workflow's
`--build-for` restores nightly snap publishing for the remaining
architectures.
Updates `go-swagger` to v0.35.0 and makes swagger generation and
validation warning-free, with the build now failing on any warning
(`go-swagger` itself exits `0` on warnings).
- Generation passes `--enable-allof-compounding` (keeps `$ref` fields
bare, no spec change) and `--skip-enum-desc` (drops the enum description
that duplicates `x-go-enum-desc` and was the only source of `allOf`
noise in the OpenAPI 3.0 output).
- Fixed warnings at the source: dropped `swagger:strfmt` where it
conflicts with `required: true` (`required` kept, `time.Time` still maps
to `date-time`), fixed a malformed `units_map` example, moved the
`parameterBodies` injection hack to `swagger:parameters`, and removed
unused responses.
Fixes: https://github.com/go-gitea/gitea/issues/12508
---------
Co-authored-by: bircni <bircni@icloud.com>
The rules from `eslint-plugin-array-func` are redundant with
`eslint-plugin-unicorn`:
- `from-map` → `unicorn/prefer-array-from-map`
- `no-unnecessary-this-arg` → `unicorn/no-array-method-this-argument`
- `prefer-flat` / `prefer-flat-map` → `unicorn/prefer-array-flat` /
`unicorn/prefer-array-flat-map` (already disabled here)
The two remaining rules (`avoid-reverse`, `prefer-array-from`) are niche
and not worth carrying an extra dependency for.
Co-authored-by: bircni <bircni@icloud.com>
- bump ci, flake and `@types/node` to node 26
- regenerate flake.lock which is needed for that package
- refactor workflow to use shared composite action
Inlines the small SVG bar graph into `RepoActivityTopAuthors.vue` (its
only consumer) and drops the `vue-bar-graph` npm dependency.
- Bars render at static height (dropped the grow animation).
- Theme-aware axis color instead of a hardcoded `#555555`.
- Removed the dangling `role="img"`/`aria-labelledby` on the `<svg>`.
- Reserve the chart height so the page does not shift when the component
mounts.
<img width="416" height="110" alt="Screenshot 2026-07-01 at 11 15 25"
src="https://github.com/user-attachments/assets/b2db4d0c-20f1-4345-9951-32a908abfaba"
/>
<img width="419" height="110" alt="Screenshot 2026-07-01 at 11 15 35"
src="https://github.com/user-attachments/assets/853305a5-575f-4a26-ba3b-12fc51081324"
/>
fyi @lafriks
---------
Signed-off-by: silverwind <me@silverwind.io>
Adds `make generate-codemirror-languages` to the npm group's
`postUpgradeTasks` in `renovate.json5`, so renovate regenerates
`assets/codemirror-languages.json` whenever `@codemirror/language-data`
(or any npm dep) updates — mirroring the existing `make svg` handling.
Also reformats the `fileFilters` arrays multi-line and regenerates the
asset to pick up current upstream linguist languages.
- Bump `eslint`, `typescript-eslint` and `eslint-plugin-unicorn` (to
v68), and configure the rules added in unicorn v66/v67/v68.
- Remove `eslint-plugin-github` and its workarounds (rules, type stub,
pnpm peer override, in-code `eslint-disable` comments); the rules worth
keeping are covered by `unicorn` equivalents.
- Apply the resulting fixes and autofixes across the JS codebase.
_Prepared with Claude (Opus 4.8)._
---------
Signed-off-by: wxiaoguang <wxiaoguang@gmail.com>
Co-authored-by: wxiaoguang <wxiaoguang@gmail.com>
Removes the legacy `delete-button` handler (`initGlobalDeleteButton`)
and migrates all remaining usages to `link-action` and `show-modal` /
`form-fetch-action`.
Two handlers are adjusted for the new request shape: webauthn key delete
reads `id` from the query, and account deletion returns `JSONError` on
validation failure.
A E2E test ist added to cover one of the use cases.
Suggested in
https://github.com/go-gitea/gitea/pull/38046#discussion_r3414936737.
---------
Co-authored-by: wxiaoguang <wxiaoguang@gmail.com>
Co-authored-by: bircni <bircni@icloud.com>
Reduces the CI cache growth and disk pressure behind the flaky `No space
left on device` failures in
https://github.com/go-gitea/gitea/issues/37974.
**`go-cache`** — the cache-seeder saved with a `restore-keys` prefix
fallback, so every `go.sum` change restored the previous cache and
re-saved the union; old module versions and stale build objects
accumulated (~3 GB → ~7 GB) and overflowed disk on smaller runners. Drop
`restore-keys` from the seeder **save** branches so each `go.sum` seeds
a clean, size-bounded cache. PR runs keep `restore-keys` for warm-start
fallback.
**`free-disk-space`** — delete the unused preinstalled toolchains in
parallel (~86 s → ~54 s) and log `df -h /` before/after.
Measured during review: the hosted `ubuntu-latest` fleet is
heterogeneous — most runners have ~89 GB free on `/` (a full pgsql
integration shard peaks at ~17 GB used), but a minority arrive nearly
full and fail mid cache-restore. The toolchain deletion is the headroom
that keeps those runners green, so it stays; the cache bound shrinks the
footprint for every runner.
Authored with assistance from Claude (Opus 4.8).
---------
Signed-off-by: silverwind <me@silverwind.io>
Co-authored-by: bircni <bircni@icloud.com>
Enable `SQLITE_JOURNAL_MODE = WAL` for the sqlite integration test
config. With modernc as the default driver, concurrent writers serialize
on SQLite's single write lock and the tail of the queue can exceed the
20s busy timeout under CI load. WAL drains the queue fast enough to stay
inside the timeout (removes rollback's fsync-per-commit and
reader-vs-commit blocking) and covers all sqlite integration tests in
one change.
---
This PR was written with the help of Claude Opus 4.7
---------
Co-authored-by: Claude (Opus 4.7) <noreply@anthropic.com>
Co-authored-by: Lunny Xiao <xiaolunwen@gmail.com>
Co-authored-by: Giteabot <teabot@gitea.io>
Two test-only changes that cut the `-race` backend unit job's critical
path, with no behavior change.
- **`modules/auth/password/hash`** — `TestHashing`/`TestVectors`
exercised the CPU-bound KDFs (scrypt `N=65536`, pbkdf2, bcrypt, argon2)
serially on one core. Marking the subtests `t.Parallel()` fans them
across cores. The hasher registry they read is only mutated by the
non-parallel `Test_registerHasher`, so this is race-free.
- **`services/release`** — `TestRelease_Update`/`TestRelease_createTag`
slept `6x time.Sleep(2s)` only to cross the 1-second `CreatedUnix`
boundary. Replaced with an advancing mocked clock (`timeutil.MockSet`),
making the timestamp assertions deterministic and removing the real
waits.
---
This PR was written with the help of Claude Opus 4.8
Co-authored-by: Claude (Opus 4.8) <noreply@anthropic.com>
Co-authored-by: Lunny Xiao <xiaolunwen@gmail.com>
Updates `github.com/urfave/cli/v3` to
[v3.9.0](https://github.com/urfave/cli/releases/tag/v3.9.0) and removes
the renovate pin now that
[urfave/cli#2319](https://github.com/urfave/cli/pull/2319) (the `-c`
help flag parsing fix) is merged.
v3.9.0 prepends the default command name to the root command's args,
which broke the old `Root().Args()` check in `isValidDefaultSubCommand`.
It now uses the command's own `Args()`.
Behavior change: `./gitea web <extra-positional-arg>` now errors with
`unknown command` instead of starting the web server and ignoring the
trailing arg. `web` takes no positional args, so this is stricter (and
arguably more correct) input handling. The intended `./gitea bad-cmd`
rejection is unchanged.
---
This PR was written with the help of Claude Opus 4.7
---------
Co-authored-by: Claude (Opus 4.7) <noreply@anthropic.com>
Co-authored-by: Nicolas <bircni@icloud.com>
Removes redundant/obsolete WebKit prefixes:
- `-webkit-mask-*` — duplicate the unprefixed `mask-*` siblings already
present; every supported browser handles unprefixed CSS Masking
longhands.
- `-webkit-overflow-scrolling: touch` — a no-op outside iOS Safari <13.
Browser floor (all support unprefixed `mask`): Chrome 120+, Safari
15.4+, Firefox 53+, and PaleMoon/Goanna (verified: unprefixed `mask`
longhands implemented unconditionally in UXP).
---
This PR was written with the help of Claude Opus 4.8
Co-authored-by: Claude (Opus 4.8) <noreply@anthropic.com>
At a 512m heap the CI Elasticsearch GC-thrashes under the jobs' memory
pressure and goes unresponsive, flaking `test-unit` (ES indexer tests
time out) and `test-mysql` (the ES-backed issue indexer blocks the
per-test queue flush). Raise the heap to 1g and disable ML + the startup
GeoIP download.
Co-authored-by: Claude (Opus 4.8) <noreply@anthropic.com>
Bump the pinned `giteabot` action to the
[`v1.0.3`](https://github.com/go-gitea/giteabot/releases/tag/v1.0.3)
release in both `giteabot.yml` and `giteabot-backport.yml`. v1.0.3 moves
label/state queries off the search API on top of the existing retry
logic.
---
This PR was written with the help of Claude Opus 4.8
Co-authored-by: Claude (Opus 4.8) <noreply@anthropic.com>
In dev mode `/api/swagger` returned HTTP 500 (`Failed to locate local
path for managed asset URI: css/swagger.css`): the backend synthesised
asset keys from the Vite entry name instead of reading the manifest,
which only worked by coincidence and broke once a source file name
diverged from its entry name.
This keys the manifest by its source path (e.g. `web_src/js/index.ts`)
and resolves entries directly — hashed `file` in prod, dev-server source
in dev. A new `AssetCSSLinks` helper renders a JS entry's stylesheet
`<link>` tags from the manifest (the entry's CSS plus the CSS of its
statically-imported chunks).
Fixes: https://github.com/go-gitea/gitea/issues/37830
Fixes: https://github.com/go-gitea/gitea/pull/37832
Fixes: https://github.com/go-gitea/gitea/pull/37876
Signed-off-by: silverwind <me@silverwind.io>
Co-authored-by: prakhar0x01 <prakharporwal2004@gmail.com>
Co-authored-by: Nicolas <bircni@icloud.com>
Co-authored-by: Claude (Opus 4.7) <noreply@anthropic.com>
Co-authored-by: Giteabot <teabot@gitea.io>
After posting a comment the page reloads via fetch-action. Clicking
"Close Issue" before the form re-initializes triggers a native form
submit, which navigates to the raw JSON redirect response
(`{"redirect":...}`) instead of the issue, so "Reopen Issue" never
appears and the test times out (observed on Firefox in CI).
Wait for the comment button to become disabled — which only happens once
the form re-initializes — before clicking "Close Issue".
---
This PR was written with the help of Claude Opus 4.7
Co-authored-by: Claude (Opus 4.7) <noreply@anthropic.com>
Co-authored-by: Nicolas <bircni@icloud.com>
- Update `github.com/alecthomas/chroma/v2` to `v2.25.0`.
- Migrate `github.com/dlclark/regexp2` to `/v2` (incorporates
https://github.com/go-gitea/gitea/pull/37664); drop the renovate pin.
- Replace the unmaintained `github.com/dimiro1/reply` (the last consumer
of `regexp2` v1 in our own code) with a small built-in reply parser for
incoming mail.
Signed-off-by: wxiaoguang <wxiaoguang@gmail.com>
Co-authored-by: Claude (Opus 4.7) <noreply@anthropic.com>
Co-authored-by: wxiaoguang <wxiaoguang@gmail.com>
Co-authored-by: Giteabot <teabot@gitea.io>
Co-authored-by: Nicolas <bircni@icloud.com>
The modernc SQLite driver (default since
https://github.com/go-gitea/gitea/pull/37562) returns `SQLITE_BUSY` once
the busy timeout is reached, unlike mattn which waited indefinitely.
`TestResourceIndex` fires many concurrent `NewIssue` writers, but SQLite
serializes all writers, so they queue on a single `BEGIN IMMEDIATE`
write lock. Under `-race` (modernc is much slower) the goroutines at the
back of the queue exceeded the hardcoded 5s test timeout, producing
`database is locked (5) (SQLITE_BUSY)`.
Changes:
- Reduce the concurrent inserts from 25 to 10. Since SQLite serializes
writers, the extra goroutines only deepen the busy-lock queue without
adding coverage. 10 still exercises concurrent index allocation while
cutting the test's `-race` runtime ~3x (2.76s to 0.86s locally).
- Share the busy-timeout constant: export `DefaultSQLiteBusyTimeout`
(20s, the production default) and reference it from the test engine
instead of the hardcoded `5000`.
Observed flake:
https://github.com/go-gitea/gitea/actions/runs/26394082930/job/77690496092
---
This PR was written with the help of Claude Opus 4.7
---------
Co-authored-by: Claude (Opus 4.7) <noreply@anthropic.com>
`TestAPIRepoMigrate` migrated from
`https://github.com/go-gitea/test_repo.git`, so it required internet
access, was slow, and could hit GitHub rate limits.
It now clones a local fixture repo (`user2/repo1`) served by the
`onGiteaRun` test server, split into two subtests:
- `Permitted` (`AllowLocalNetworks=true`) — the success/permission
cases, cloning the local repo.
- `DisallowedHost` (`AllowLocalNetworks=false`) — the private-IP
rejection cases.
The split is needed because those two settings are mutually exclusive.
The clone address is built from the live listener (`u`) so it can't
drift from the bound host/port. The permission matrix and
disallowed-host assertions are unchanged.
Test is now roughly 2.5 times as fast with while asserting the same as
before without a GitHub dependency.
---
This PR was written with the help of Claude Opus 4.7
Co-authored-by: Claude (Opus 4.7) <noreply@anthropic.com>
Co-authored-by: Nicolas <bircni@icloud.com>
Reduces CI minutes consumption by narrowing the `files-changed` filters.
- DB matrix (`pgsql`/`mysql`/`mssql`/`unit`) now runs only on real
backend changes. `test-sqlite` stays gated on `actions`, so it remains
the smoke check that validates CI-infra changes (composite-action edits,
workflow edits, renovate action-pin bumps) without spinning up the full
matrix.
- Fix the `templates` filter: the SVG template linter is
`tools/lint-templates-svg.ts`, so the `tools/lint-templates-*.js` glob
matched nothing.
- Add missed paths: `tsconfig.json` and
`tools/generate-svg-vscode-extensions.json` to `frontend`,
`eslint.json.config.ts` to `json`, and
`.github/actions/docker-dryrun/**` to `docker`.
---
This PR was written with the help of Claude Opus 4.7
---------
Signed-off-by: silverwind <me@silverwind.io>
Co-authored-by: Claude (Opus 4.7) <noreply@anthropic.com>
Elasticsearch JVM heap defaults to ~50% of detected host RAM, typically
way too much for our little tests and it starved the other runner
processes from memory.
Co-authored-by: Claude (Opus 4.7) <noreply@anthropic.com>
- Adds `make lint-shell`. uses local `shellcheck` if its version
matches, otherwise runs the pinned image via docker or podman
- `.shellcheckrc` disables the most annoying rules
- Fixes findings across existing scripts
Fixes: #37648
Signed-off-by: silverwind <me@silverwind.io>
Co-authored-by: Claude (Opus 4.7) <noreply@anthropic.com>
Co-authored-by: TheFox0x7 <thefox0x7@gmail.com>
Co-authored-by: Nicolas <bircni@icloud.com>
1. Split the psql matrix jobs into composite actions. Matrix jobs that
can skip do not work with required checks on GitHub because skipped and
unskipped emit different job names (GitHub bug
https://github.com/orgs/community/discussions/9141).
2. Dedupe node and go setup steps into composite actions
Currently test-psql branch protection is disabled, will re-enable when
merging this.
---------
Signed-off-by: silverwind <me@silverwind.io>
Co-authored-by: Claude (Opus 4.7) <noreply@anthropic.com>
Co-authored-by: Nicolas <bircni@icloud.com>
Adds Playwright e2e coverage for five high-value workflows, each driven
through semantic locators with API-based setup:
- comment on and close an issue
- publish a release
- star and watch a repository
- create a pull request from the compare page
- fork a repository
Also passes `autoInit: false` in existing tests that only exercise
DB-backed units (issues, reactions, milestones, projects, events),
skipping an unused initial commit to speed up their setup and reduce
parallel git contention.
---
This PR was written with the help of Claude Opus 4.7
---------
Co-authored-by: Claude (Opus 4.7) <noreply@anthropic.com>
Co-authored-by: Nicolas <bircni@icloud.com>
Critical path ~25:42 → ~19:56 (−22%), ~0% CI minutes.
- `test-pgsql` shards 2-way. Branch protection: replace `test-pgsql`
with `test-pgsql-shards (1)` + `test-pgsql-shards (2)`; `test-unit`,
sqlite/mysql/mssql unchanged — pgsql dominates the critical path.
- `test-unit` runs `bindata` then `bindata gogit` sequentially.
cache-seeder pre-warms the race-instrumented test compile cache and the
integration test binary so PR jobs warm-start.
- Cache writes restricted to cache-seeder; PR jobs use
`actions/cache/restore`. Defends against PR cache poisoning and frees
the 10 GB cap from PR churn.
- `go-cache` action: dropped the `cache-name` input. One gobuild cache,
one golangci-lint cache. Seeder lint job restores but doesn't save
gobuild, so only one writer populates it.
- `tools/test-integration.sh` shards the integration binary via
`-test.list`; `TestMain` short-circuits DB init in list mode.
`TestAPILFSNotStarted` / `TestAPILFSLocksNotStarted` switched to
`test.MockVariableValue` — latent `setting.LFS.StartServer` global-state
leak uncovered by sharding.
---
This PR was written with the help of Claude Opus 4.7
---------
Signed-off-by: silverwind <me@silverwind.io>
Co-authored-by: Claude (Opus 4.7) <noreply@anthropic.com>
Co-authored-by: wxiaoguang <wxiaoguang@gmail.com>
Co-authored-by: Nicolas <bircni@icloud.com>
Indented (4-space) code blocks were emitted by goldmark's default
renderer as plain `<pre><code>` without the `code-block-container`
wrapper that the JS `initMarkupCodeCopy` keys on. As a result, only
fenced code blocks received the copy button. Register
`ast.KindCodeBlock` with a renderer that produces the same wrapper as
the highlighting renderer so both syntaxes get the button.
Extends `TestMarkdownFencedCodeBlock` to assert the wrapper is emitted
for indented blocks (and that HTML inside is escaped).
Co-authored-by: Claude (Opus 4.7) <noreply@anthropic.com>
Co-authored-by: wxiaoguang <wxiaoguang@gmail.com>
Fixes two recurring CI failures:
1. `cache-seeder.yml` lint-backend missing a `make generate-go` before
linting with `TAGS=bindata`, and `pull-compliance.yml` lint-on-demand
failing its post-step pnpm cache save when no pnpm-using conditional
step runs.
2. Drops `cache: pnpm` from lint-on-demand and adds `make generate-go`
to cache-seeder's lint job.
---
This PR was written with the help of Claude Opus 4.7
---------
Signed-off-by: silverwind <me@silverwind.io>
Co-authored-by: Claude (Opus 4.7) <noreply@anthropic.com>
Drop the broad `tools/*.{js,ts}` globs from the `frontend` filter so
edits to CI-only or backend helper scripts in `tools/` stop triggering
frontend and e2e jobs. Only `tools/generate-svg.ts` is kept.
Also renames `tools/lint-pr-title.js` to `.ts` for consistency, drops
the empty root `*.js` glob, fixes stray indentation in the `dockerfile`
filter and adds missing `setup-node`.
---
This PR was written with the help of Claude Opus 4.7
---------
Co-authored-by: Claude (Opus 4.7) <noreply@anthropic.com>
The truncated `ActionRunner` struct in
`AddCancellingSupportToActionRunner` declares only the new
`HasCancellingSupport` column. When xorm's `SyncWithOptions` compares it
against the live `action_runner` table, every index/constraint absent
from the local struct is a candidate for removal.
Walking [xorm v1.3.11
sync.go:250-266](https://gitea.com/xorm/xorm/src/tag/v1.3.11/sync.go#L250-L266):
- `IndexType` indices skip the drop when `IgnoreIndices ||
IgnoreDropIndices` — already covered.
- `UniqueType` indices skip the drop only when `IgnoreConstrains` —
**not** set in #37275, so the existing `UNIQUE` on `token_hash` (and any
other uniques) would be dropped on upgrade.
Adding `IgnoreConstrains: true` matches v333's pattern and preserves the
existing unique constraints. Spotted by @wxiaoguang in
https://github.com/go-gitea/gitea/pull/37275#discussion_r3254168680.
---
This PR was written with the help of Claude Opus 4.7
Co-authored-by: Claude (Opus 4.7) <noreply@anthropic.com>
Co-authored-by: Nicolas <bircni@icloud.com>
`act/workflowpattern` in runner is only consumed by Gitea and dead code
there. Move it to this repo. Use `modules/glob` for glob pattern match.
---------
Co-authored-by: Claude (Opus 4.7) <noreply@anthropic.com>
Co-authored-by: wxiaoguang <wxiaoguang@gmail.com>
Adds [zizmor](https://docs.zizmor.sh/) to `make lint-actions` with
`--min-confidence=medium`. Fixes the remaining findings:
- Pin floating-tag service images in `pull-db-tests.yml` to
`tag@sha256:digest`
- Move `github.ref` / `github.ref_name` (and surrounding secrets/step
outputs for consistency) out of `run:` into `env:`
---
This PR was written with the help of Claude Opus 4.7
---------
Signed-off-by: silverwind <me@silverwind.io>
Co-authored-by: Claude (Opus 4.7) <noreply@anthropic.com>
Co-authored-by: Nicolas <bircni@icloud.com>
Co-authored-by: Giteabot <teabot@gitea.io>
Enables Renovate's `dockerfile` manager so the base images in
`Dockerfile` and `Dockerfile.rootless`
(`docker.io/library/golang:1.26-alpine3.23`,
`docker.io/library/alpine:3.23`) get tracked alongside the other
dependencies. Updates are grouped under "docker dependencies" and follow
the existing weekly schedule.
---
This PR was written with the help of Claude Opus 4.7
Co-authored-by: Claude (Opus 4.7) <noreply@anthropic.com>
Co-authored-by: Nicolas <bircni@icloud.com>
Add two rules to `AGENTS.md` for recurring issues.
Signed-off-by: silverwind <me@silverwind.io>
Co-authored-by: Claude (Opus 4.7) <noreply@anthropic.com>
